Chatbot Design: 10 Things to Keep in Mind

Photo by UX Indonesia on Unsplash

Have you ever come across chatbots that are monotonous, ask the same information again and again, or work only when a predefined flow is followed? Frustrating, right? What do you think might be the reason behind this? — — Yes, poor design!

We can never expect the user to behave predictably or just ask the questions that the bot has been trained to answer. No matter how well the bot is developed, without a proper design and planning phase, it can permanently damage your brand image and alienate a potential customer.

Are Chatbots Really the Future?

Here are 10 things to keep in mind before proceeding to the development phase:

1. Define the use-case
   The first step before the development phase is to analyze the requirements and carefully define the use case. Define the scope of the chatbot, what the domain is and what questions need to be answered.
   “Will a chatbot solve the problem at hand?”
   “How much will it add in terms of value?”
2. Set clear expectations
   Unlike a Virtual Assistant, a chatbot, which is here to solve a specific problem does not need to answer all the queries of the customer. Therefore setting clear expectations and making customers aware of the scope of the bot is very essential. Also, be transparent and let the users know that they are conversing with a chatbot.
3. Drive the conversation
   The next step is to design the conversational flow. Most chatbots simply respond to a user query and then wait for the user to ask a question again. Rather than that, chatbots could be smart enough to drive the conversation and keep it flowing with follow-up prompts and cue’s whenever needed, thereby increasing user engagement.
4. Interactive content
   Rather than mundane text, interactive content often makes the conversation more interesting and encourages people to chat more often with the chatbot. Make sure you add images, emojis, gifs, and videos along with properly formatted text responses as much as possible.
5. Bot persona
   Your bot’s personality should be aligned with the organization’s core values. You could begin by giving a name and an image that complements your brand vision.
6. Know your audience
   Personalized conversations — Creating a conversational experience tailored to each person would help in more adoption of the solution and increased user satisfaction.
   Don’t assume
   Human beings have a lot of unconscious biases. We must ensure that our prejudices are not passed on to the bots we build and that they do not have a detrimental impact on the user experience. Take conscious efforts to avoid perpetuating gender or ethnic stereotypes. Many things could go wrong if we have, say a chatbot that interacts with a person by assuming gender.
   Some other types of biases are:
   * Interaction bias: Microsoft’s Tay bot which learned from human interactions and turned out to be racist, is one example.
   * Dataset bias: When the data used for training the model isn’t representative of the diversity of the user base.
7. Accessibility
   Accessibility is an important aspect of customer retention and satisfaction. Some of the ways in which a chatbot can be made accessible are:
   * Images should be accompanied by alternative text, and audio/videos should include closed captioning or transcripts.
   * The chatbot should be accessible by using the keyboard only.
   * Accessible language, colour, fonts should be used.
   * As far as possible, avoid jargon and acronyms.
   * Voice-only Design — Taking a step further, this design is based on creating a conversational bot that can have complete conversations with users solely through voice and without the use of textual input. Other visual elements like images, videos, or audio can enrich the experience but are not necessary for the conversation.
8. Fall gracefully
   If provided with the free text box, people often ask all sorts of queries merely to break the chatbot. In that case, it's always better to respond with a sorry than to give out a wrong response and mislead the users. Also rather than just displaying a sorry message, it's better to provide some value to the user by using the following:
   * Follow-up prompts
   Display questions that are closely relevant to the user’s question and provide users with the option to choose from them.
   * Help card/page
   This informs the users about the scope of the bot and also displays the complete capabilities of the bot.
   * Agent handoff
   Provide the option to talk to a customer support agent if the bot is unable to answer the question and also whenever the user explicitly asks for it.
9. Security
   Chatbots usually store personal data to identify the user and personalize the customer experience. Ascertain that the chatbot complies with all GDPR regulations, respects user privacy, and handles data securely. It's important to show the Privacy Policy at the outset of the conversation as well as when requested. Users should be able to access, download and delete the collected data if they desire to. Ensure that no PII(Personally Identifiable Information) is stored without user consent.
   “What data is collected?”
   “Why is it collected?”
   “How can the data be accessed?”
   “How long will it be stored?”
10. Feedback and insights
    Tracking and analyzing data can help enhance both the customer experience and the chatbot performance iteratively. Data visualization tools can help us understand trends, patterns and gain insights from the data. Usage statistics, the list of failed queries, frequently asked questions, the number of agent handovers, and other user data are good indicators of how well the chatbot performed.
    Other ways to obtain user feedback are:
    * A simple feedback form or survey after the end of the conversation.
    * Each message can contain a thumbs up and thumbs down emoji as optional feedback buttons.
    “How is the success of the chatbot measured?”
    "How can we gain a deeper understanding of our customers?”
    “How can user feedback be captured and visualized?”
    “How does the process of refinement look like?”

This is just a basic checklist, a chatbot design process is far more nuanced and iterative. I hope this helps you get started! Happy Botting!

About The Author
Geethu Suresh is a Microsoft .Net Consultant here at Version 1.